FFMPEG Transcoding Perfromance not good on Tesla p4

Hi All,
I am working on a transcoding application on Tesla P4. With reference to the the below document, http://on-demand.gputechconf.com/gtc/2017/presentation/s7111-abhijit-patait-nvidia-video-technologies.pdf
I tried transcoding multiple video files. In slide 44, perfromance benchmark it says it can do 13 4K,8bit, streams@30fps/NVENC. But my encoding numbers and the number of streams are far less compared to the numbers in document. For 6-Streams, with preset HP I am getting only 20fps/stream, which is 120fps max. As per the document it should 26, 4K streams should be supported on tesla P4. Please let me know if I am missing anything.

Below are the command line arguments I am using to encode in HEVC using NVENC.
-Encoding using YUV
ffmpeg -f rawvideo -s:v 3840x2160 -r 120 -pix_fmt yuv420p -i YachtRide_3840x2160.yuv -vcodec hevc_nvenc -preset hp -cq 10 -g 150 outputhp1.h265
-Transcoding
ffmpeg -y -vsync 0 -hwaccel cuvid -c:v h264_cuvid -i outputhp.h264 -c:a copy -vf scale_npp=1280:720 -c:v hevc_nvenc -b:v 40M output_new.h265

Thanks in advance

Hello there,

Can you show us the output of:

(a).GPU information?

nvidia-smi -q

(b). For topology information:

nvidia-smi topo -m